Data-Driven Analysis of Senior High Students’ Sentiments and Strand Selection Using Machine Learning

Abstract

Senior high school students’ sentiments play a vital role in shaping educational strategies that promote mental health, well-being, and learner protection as mandated by the Department of Education’s child-friendly school system. This study investigates the relationship between students’ sentiments and their chosen academic strand using machine learning algorithms. A total of 580 students participated, providing feedback through a guided support session conducted after the first quarter of the school year. The sentiments that were collected, which are code-mixing data, were preprocessed and vectorized using Term Frequency–Inverse Document Frequency (TF-IDF). A Support Vector Machine (SVM) classifier with four-fold cross-validation was employed to categorize sentiments into positive, neutral, and negative classes. The model achieved an average accuracy of 72.59% with a standard deviation of 3.91%, indicating a statistically significant relationship between students’ sentiments and their chosen strands. Findings suggest the need for a dedicated technical working group to monitor and interpret sentiment trends per strand, enabling timely interventions. Future research is encouraged to explore alternative machine learning algorithms, such as Multinomial Naive Bayes, Logistic Regression, and Gradient Boosting (LightGBM), as well as advanced techniques like semantic embeddings and sarcasm detection, to improve classification accuracy. The results of this study contribute to data-driven decision-making and provide actionable insights for educational policy frameworks aimed at enhancing student welfare and learning outcomes.